开放定址散列法和再散列
目录
前面利用分离链接法解决了散列表插入冲突的问题,而除了分离链接法外,还可以使用开放定址法来解决散列表的冲突问题。
开放定址法在遇见冲突情形时,将会尝试选择另外的单元,直到找到空的单元为止,一般来说,单元h0(X), h1(X), h2(x)为相继尝试的单元,则hi(X)=(Hash(X)+F(i)) mod TableSize,其中F(i)即为冲突解决的探测方法,
开放定址法中的探测方法的三种基本方式为,
- 线性探测法:探测步进为线性增长,最基本的方式为F(i)=i
- 平方探测法:探测步进为平方增长,最基本的方式为F(i)= i2
- 双散列:探测方法的步进由一个新的散列函数决定,最基本的方式为F(i)= i*Hash2(i),通常选择Hash2(i)=R-(X mod R),其中R为小于TableSize的素数。
对于使用平方探测的开放定址散列法,当元素填得太满的时候,操作运行的时间将消耗过长,而且插入操作有可能失败,此时则可以进行一次再散列来解决这一问题。
再散列会创建一个新的散列表,新的散列表大小为大于原散列表大小2倍的第一个素数,随后将原散列表的值重新散列至新的散列表中。
这一操作的开销十分大,但并不经常发生,且在发生前必然已经进行了多次插入,因此这一操作的实际情况并没有那么糟糕。
散列的时机通常有几种,
- 装填因子达到一半的时候进行再散列
- 插入失败时进行再散列
- 达到某一装填因子时进行散列
完整代码
1 from functools import partial as pro 2 from math import ceil, sqrt 3 from hash_table import HashTable, kmt_hashing 4 5 6 class RehashError(Exception): 7 pass 8 9 10 class OpenAddressingHashing(HashTable): 11 def __init__(self, size, hs, pb, fn=None, rf=0.5): 12 self._array = [None for i in range(size)] 13 self._get_hashing = hs 14 self._hashing = hs(size) if not fn else fn 15 self._probing = pb 16 self._rehashing_factor = rf 17 18 def _sniffing(self, item, num, hash_code=None): 19 # Avoid redundant hashing calculation, if hashing calculation is heavy, this would count much. 20 if not hash_code: 21 hash_code = self._hashing(item) 22 return (hash_code + self._probing(num, item, self.size)) % self.size 23 24 def _get_rehashing_size(self): 25 size = self.size * 2 + 1 26 while not is_prime(size): 27 size += 1 28 return size 29 30 def rehashing(self, size=None, fn=None): 31 if not size: 32 size = self._get_rehashing_size() 33 if size <= (self.size * self.load_factor): 34 raise RehashError('Rehash size is too small!') 35 array = self._array 36 self._array = [None for i in range(size)] 37 self._hashing = self._get_hashing(size) if not fn else fn 38 self.insert(filter(lambda x: x is not None, array)) 39 40 def find(self, item): 41 hash_code = ori_hash_code = self._hashing(item) 42 collision_count = 1 43 value = self._array[hash_code] 44 45 # Build up partial function to shorten time consuming when heavy sniffing encountered. 46 collision_handler = pro(self._sniffing, hash_code=ori_hash_code) 47 48 while value is not None and value != item: 49 hash_code = collision_handler(item, collision_count) 50 value = self._array[hash_code] 51 collision_count += 1 52 return value, hash_code 53 54 def _insert(self, item): 55 if item is None: 56 return 57 value, hash_code = self.find(item) 58 if value is None: 59 self._array[hash_code] = item 60 if self.load_factor > self._rehashing_factor: 61 self.rehashing() 62 63 64 def is_prime(num): # O(sqrt(n)) algorithm 65 if num < 2: 66 raise Exception('Invalid number.') 67 if num == 2: 68 return True 69 for i in range(2, ceil(sqrt(num))+1): 70 if num % i == 0: 71 return False 72 return True 73 74 75 def linear_probing(x, *args): 76 return x 77 78 79 def square_probing(x, *args): 80 return x**2 81 82 83 def double_hashing(x, item, size, *args): 84 r = size - 1 85 while not is_prime(r): 86 r -= 1 87 return x * (r - (item % r)) 88 89 90 def test(h): 91 print('\nShow hash table:') 92 h.show() 93 94 print('\nInsert values:') 95 h.insert(range(9)) 96 h.show() 97 98 print('\nInsert value (existed):') 99 h.insert(1) 100 h.show() 101 102 print('\nInsert value (collided):') 103 h.insert(24, 47) 104 h.show() 105 106 print('\nFind value:') 107 print(h.find(7)) 108 print('\nFind value (not existed):') 109 print(h.find(77)) 110 111 print('\nLoad factor is:', h.load_factor) 112 113 114 if __name__ == '__main__': 115 test(OpenAddressingHashing(11, kmt_hashing, linear_probing)) 116 print(30*'-') 117 test(OpenAddressingHashing(11, kmt_hashing, square_probing)) 118 print(30*'-') 119 test(OpenAddressingHashing(11, kmt_hashing, double_hashing))